Are you ready for Windows 7?
“Hold the line!” That’s the new rallying cry for the core Windows development team. Add new features. Tweak existing ones. But whatever you do, don’t make Windows 7 any slower or fatter than Vista.
I have little doubt those are the marching orders for Windows 7, given the tight release timeframe of 18 to 24 months plus various reports of early Milestone builds. More ambitious changes would risk another — and potentially fatal — Longhorn-style delay. Windows 7 will be exactly what the internal Microsoft working title conveys: the seventh (actually fifth) generation of the Windows NT code base — the same code base that forms the basis of Windows XP and Vista today.
Shocking? Only if you’re one of the deluded Save XP die-hards who bought into the whole Windows 7 uber alles mystique. For these lost souls, the next Windows is more than just another version. It’s a true panacea, a conduit through which they can pour all of their anti-Vista angst. Don’t like UAC? Windows 7 will fix that. Frustrated by Vista’s sluggish performance? Windows 7 will run circles around it.
Reality check: Windows 7 will be a lot like Vista. In fact, it’ll be more like an extensive Service Pack (think Windows XP SP2 and/or the various NT Option Packs of yore) than a major new release. Big ideas and big new features are what got Microsoft into the whole “Longhorn reset” mess in the first place.
This is actually a good thing. Despite the criticisms leveled against it (including more than one heated diatribe by yours truly), Vista isn’t really flawed in any fundamental way. Yes, it’s slower than XP — but that was to be expected given its more complex code paths. Likewise, the “girth” issues were somewhat inevitable. Meanwhile, the hardware base is slowly catching up to where it needs to be to support a more complex Windows OS.
I’d even go so far as to say that, if Vista were launched today –- with the SP1 tweaks and improved device driver ecosystem in place -– it would fare a lot better than it did. But hindsight is 20/20. The future, in the form of Windows 7, is all about shipping an incremental follow-on to Vista that shores up the NT code base once and for all.
The good news is that this also makes speculating about the next version’s runtime behavior a lot easier. After all, if Windows 7 is just Windows Vista with some performance and usability tweaks, it means we can deduce a lot about the product’s system requirements and compatibility with the installed base by examining performance and usage data collected from systems running its immediate predecessors, Windows “5″ (also known as XP) and “6″ (also known as Vista).
Peering into the future with Windows Sentinel
Enter the Windows Sentinel project. With nearly 2,000 contributing systems, the exo.repository –- which is the heart of Windows Sentinel –- provides us with a representative sample of Windows-based systems running a mixture of versions (XP, 2003, 2008, Vista) and workloads (business productivity, analytics, home/personal).
Basically, we have our finger on the pulse of the Windows landscape. And by measuring that pulse, plus a few other metrics (and some educated guessing), we can tell a lot about how Windows 7 will be received when it ships.
For example, we can tell right now that roughly 29 percent of current systems will be able to run Windows 7, although not always with adequate performance.
This conclusion is based on an analysis of system disclosure data (CPU type/count/speed, memory size) as well as performance indices calculated from runtime data collected over a one-month period. Of the remaining systems, 60 percent have too little memory (less than 2GB) to reasonably host a Vista successor, while 29 percent don’t have the CPU horsepower (less than 2GHz).
We can further break down the “survivors” by analyzing their current workloads. A full 36 percent of them are already CPU bound, with 27 percent of them heavily overloaded. This is per the exo.repository’s Peak CPU Saturation Index, which is a compound index derived from 4 contributing factors: Processor Queue Length; Per-process Instant Delay (a custom CPU metric derived from the Processor Queue value); Per-Process Cumulative Delay (another custom metric); and Event Duration.
Interestingly, of the systems that are most heavily burdened, only 31 percent are running Vista (which is not really surprising since Vista makes up just 16 percent of the overall sample set). The rest are running a mixture of XP and Server 2003/XP-64-bit.
Needless to say, a heavily loaded (in terms of CPU saturation) XP box doesn’t bode well for Vista’s successor. Like Vista, Windows 7 will introduce a much higher CPU burden in the form of additional background services and their corresponding execution threads. Currently, this ratio runs approximately two to one in favor of Vista: A Vista-based PC must juggle roughly twice as many concurrent execution threads as an XP PC while running the identical office productivity workload.
Chances are good that Windows 7 will, at minimum, maintain the status quo in terms of resource requirements. This, in turn, means that customers who were hoping for some relief with Windows 7 will be sorely disappointed. It’s simply not realistic to expect Microsoft to produce a “leaner” OS and yet still add enough functionality to make it worth upgrading. At best, we might see a new version with a resource footprint similar to Windows Vista, which still places it out of reach of more than three-quarters of the systems in our sample set.
Bottom line: Less than 20 percent of the installed base is ready to migrate to Windows 7 today based on all of the factors detailed above. By far, the biggest (60 percent of the base) inhibitor is limited RAM: Like Vista, Windows 7 will have a voracious appetite for memory. Today, 2GB is the bare minimum for reasonable Vista performance. Expect 4GB to be the norm by the time Windows 7 ships. Likewise, the days of the single-core CPU are over. Dual cores are a must, while quad cores are rapidly transitioning from luxury to mainstream necessity.
Fortunately for Microsoft, it has time –- and Moore’s indefatigable Law –- on its side. Assuming Microsoft does indeed “hold the line” on code path expansion and keep Windows 7′s requirements within striking distance of Windows Vista, it can launch virtually anything and still have a winner. Just don’t expect to boot it on that old Pentium III box you stumbled across in your basement.
Posted in Hardware, Microsoft, Software | No Comments »
